vllm
https://github.com/vllm-project/vllm
Python
A high-throughput and memory-efficient inference and serving engine for LLMs
Triage Issues!
When you volunteer to triage issues, you'll receive an email each day with a link to an open issue that needs help in this project. You'll also receive instructions on how to triage issues.
Triage Docs!
Receive a documented method or class from your favorite GitHub repos in your inbox every day. If you're really pro, receive undocumented methods or classes and supercharge your commit history.
Python not yet supported59 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [RL][Model Loader] Direct reload: check that an update loaded every parameter
- [Model Loader] Direct reload of online-quantized modules
- [RL][Model Loader] Load plan: stacked (transformers v5) routed experts as per-expert views
- [KV Events] Add opt-in inline event source attribution
- [Quantization] Add a KV-cache KL benchmark that reads the quantized cache
- [Bugfix][Attention] Zero-init dense attention output buffers to prevent CUDA-graph padding corruption (#65734)
- [Bug]: FP8 KV cache with `kv_cache_dtype_skip_layers` crashes FA3: "scheduler_metadata must have shape (metadata_size)"
- [Perf][KV Connector][ROCm] MoRIIO WRITE: skip blocks the decoder prefix cache already holds
- [Model] Add LFM3-Omni
- [Bugfix][Attention] Prevent overlapping query tile writes in Triton unified attention
- Docs
- Python not yet supported